Cure is a significant process during back end of the line fabrication of integrated circuits with hydrogen silsesquioxane since it affects structure and properties of the spin on dielectric material. Reported herein is the effect of soak temperature, time, and oxygen concentration process parameters on structure and properties of hydrogen silsesquioxane. Results of the study emphasize the importance of an inert environment during the baseline recommended cure conditions of 400 °C for one hour in order to avoid oxidation and formation of polar silanol or water species. A 350 °C cure temperature is more robust to oxidation providing similar or improved properties. Shorter cure times result in similar structure and properties as the baseline cure which suggests that lower temperature and/or shorter cure time may provide value worth investigating by integrated circuit manufacturers.
IntroductionFiber-optic communication systems require large number of channels (DWDM) and high signal power to meet the demands of fastly growing data traffic and longer link spans. It is well known that the transmission capacity of DWDM systems is limited by such fiber nonlinearities as four-wave mixing (FWM), cross-phase modulation (XPM) and self phase modulation (SPM). Nonlinearity in EDFAs has not been a major concern until recently. Eiselt el. al. [l-21 investigated the impact of XPM in both C-and L-band EDFAs and showed that it is could be comparable with and even exceed that of a transmission fiber for DWDM systems. Thus, nonlinear processes in Erbium fiber are important for DWDM system perfonnnce.In this paper, we experimentally and theoretically investigate FWM in both C-and L-band EDFAs. We found that FWM cross-talk level in L-band could significantly exceed that in C-band, which can be explained by a longer coil length in L-band. We show that FWM efficiency is proportional to the square of the length of the Er-doped fiber (EDF) and FWM efficiency in EDFA has a "weak" scaling with channel spacing unlike that of transmission fiber. To evaluate the impact of EDFA-induced FWM on DWDM system performance, we analyze 5xlOOkm. 40xlOGb/s, DWDM NU-format system in Lband and show that it could degrade system Q by up to 1 dB. ExperimentThe experimental schematic is shown in Fig. 1. We use two tunable CW laser sources (TLS), the polarization of which is aligned by polarization controllers. The two input signals then pass through a polarizer to ensure co-polarization of signals for maximum FWM efficiency. The signals are launched into a single-coil EDFA, with a forward 980 pump and a backward 1480 pump. The single-coil EDF used in this experiment is a high concentration EDF with a peak absorption of 20 dB/m at 1530 nm. The input power per channel is in the range of 1 dBm to 2.5 dBm and the output power per channel is in the range of 13 dBm to 15 dBm. The length scale for the L-band and C-band operation are obtained by scaling the effective gain coefficient (a ratio of 5). An optical spectrum analyzer (OSA) monitors the output power spectra. Fig. 2 shows the output spectra of the two signals and their FWM mixing tones in both C-and Lband. The two signal channels (PI and P2) are spaced by 1 nm in this case. Clearly, the FWM power is much lower in C-band than L-band. Shown in Fig.3 is the FWM efficiency in both C-band and L-band as a function of channel spacing. For the C-band measurement, PI is at 1555 nm and P2 is tuned away from this channel. For the L-band measurement, PI is at 1580 nm and P2 is tuned way from this channel. The FWM efficiency (qWM in dB) is defined in a standard way as PWM-2P2-PI (dB).First, we can see a reduction of FWM efficiency by 14 dB in C-band as compared to L-band. This can be explained as a result of a length decrease by a factor of 5 in C-band. Second, FWM efficiency in both L-band and especially C-band decreases with the channel spacing AA much slower compared with what it should be acco...
Multiwavelength SystemsOver the last four years most of the current North American terrestrial long-haul fiber network has been upgraded to operate at 1.7 to 2.4 Gbfs. Long term trends indicate that fiber data rates double every 2 years. The introduction of 10 Gb/s systems is expected in 1996 and we can predict that there will be a need for 40 Gb/s by the end of the decade.In the past, increasing the data rate (TDM) has been the method of choice to reach higher fiber capacities, but now systems that use multiple wavelength channels (WDM) at both 2.4 and 10 Gb/s are being considered [ 11. This change has been driven by a number of factors the most critical of which has been the introduction of the Er-doped fiber amplifier (EDFA). A single EDFA, in contrast to an optoelectronic repeater, can ampllfy multiple wavelength channels. Secondly, the maximum fiber dispersion that can support a given data rate distance product goes down as the square of the data rate. Most of the installed fiber network is unshifted fiber which has a dispersion of between 15-20 ps/nm km in the 1550 nm window. With this dispersion, 2.4 Gb/s can be carried over 500 km, but 10 Gb/s transmission will require dispersion compensation. At 40 Gb/s the dispersion tolerance will be so small that dispersion compensation may even be required for dispersion-shifted fiber. Also, in a TDM system at an add-drop point, all of the traffic must be processed and the add-drop must operate at the trunk-line rate. For a WDM add-drop, only one wavelength need be dropped while the bulk of the traffic is untouched and the add-drop electronics need only operate at the individual channel rate. Finally, in a WDM system the upgrade to higher fiber capacity can be gradual as individual wavelengths are added only as needed on a particular route. Multiwavelength AmplifiersThe shape of the gain spectrum is critical for a multiwavelength amplifier. The S N R of the low gain channels w i l l limit the number of amplifiers that can be cascaded. In Er-Al-silica, a reasonably flat gain spectrum can be obtained between 1540 and 1560 nm. To go below 1540 nm requires a gain shaping filter. Other requirements on the amplifier are: high gain, to support long spans between amplifiers, efficient use of pump, so that many wavelength channels can be camed, and low noise, so that the amplifers can be cascaded. The combination of all of these requirements favors 980 nm pumping and requires the use of multi-stage amplifier designs. We have built a multi-stage amplifier for multi-channel amplification between 1549 and 1561 nm. Over that range there is less than a dB of gain variation with 34 dB gain. Operating with 34 dB gain, 72% of the pump power delivered to the gain fibers is converted into signal, and the noise figure is 4.2 dB.Numerical modeling can be used to study the behavior of multiwavelength amplifier cascades. Figure 1 shows the predicted optical path Q, in dB, for each of 8 channels at the end of a 500 km system with 4 amplifiers. The Q values are all above 25 dB and they vary ...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.